DeepSeek API: Models, Pricing, and How to Call It (2026)

The DeepSeek API in One Paragraph

DeepSeek is an AI lab that ships open-weight Mixture-of-Experts models and a hosted API to call them. The API endpoint is api.deepseek.com and it speaks the OpenAI wire format, so any OpenAI SDK works after you change the base URL and model name. Two models are live: deepseek-v4-flash for cost-sensitive workloads and deepseek-v4-pro for higher capability.

Both API models carry a 1M-token context window. Pricing is per million tokens and split by whether the input prefix is a cache hit or a cache miss. A cache hit (a repeated system prompt or few-shot block already in DeepSeek's context cache) is priced 50x lower than a cache miss on flash.

The open-weight side is where the architecture lives. DeepSeek-V3.2-Exp is 685B total MoE parameters under an MIT license, and DeepSeek-R1 is 671B total with 37B activated. These weights are downloadable for self-hosting, which is the main reason DeepSeek shows up in cost comparisons against closed APIs.

DeepSeek Models and Pricing

The table below lists the DeepSeek API models with context, per-1M input pricing (cache miss), per-1M output pricing, and notes. All figures are from DeepSeek's pricing page as of June 2026.

deepseek-v4-flash is the default for most work. At $0.14/1M input and $0.28/1M output it is one of the cheapest 1M-context models with frontier-adjacent coding scores. deepseek-v4-pro costs roughly 3.1x the input and 3.1x the output of flash and is the choice when flash misses on harder reasoning.

Cache Hit vs Cache Miss Pricing

DeepSeek prices input tokens differently depending on whether the prefix is already in its context cache. A cache hit means a leading chunk of your prompt (system instructions, few-shot examples, a long document you keep re-sending) matches a recently processed prefix. DeepSeek charges the cache-hit rate for those tokens and the cache-miss rate for the rest.

The practical effect is large for agent workloads. A coding agent re-sends the same system prompt and tool definitions on every turn. Those tokens land as cache hits, so the marginal input cost of each turn drops toward the cache-hit rate. Structure prompts so the stable prefix comes first and the per-request content comes last to maximize cache hits.

The tradeoff: cache behavior is not guaranteed across every request, and pricing assumes a recently seen prefix. Cold prefixes, frequently changing system prompts, and one-shot calls pay the full cache-miss rate. Do not budget your whole input at the cache-hit price.

deepseek-chat vs deepseek-reasoner

deepseek-chat answers directly. deepseek-reasoner emits a chain-of-thought before the answer, which raises accuracy on math and multi-step coding at the cost of more output tokens and higher latency. Both are modes of the same underlying deepseek-v4-flash model and both legacy names retire on 2026/07/24.

Because reasoner produces more output tokens, its effective cost per answer is higher even at the same $0.28/1M output rate. Reserve thinking mode for prompts where a chain-of-thought measurably improves the result, and route easy prompts to non-thinking. A router that classifies difficulty handles this split automatically, which is the same pattern Morph's LLM router applies across providers.

Context Window and Architecture

The hosted API models expose a 1M-token context window. The open-weight checkpoints behind the line expose smaller windows: DeepSeek-V3.2-Exp supports 163,840 tokens (160K) and DeepSeek-R1 supports 128K. If you self-host the weights, plan around the checkpoint window, not the 1M API figure.

DeepSeek-V3.2-Exp introduces DeepSeek Sparse Attention (DSA), a sparse attention mechanism for long-context training and inference efficiency, per its Hugging Face model card. Sparse attention is what makes a 160K-token window practical on a 685B-parameter MoE without the quadratic cost of dense attention at that length.

DeepSeek-R1 is a Mixture-of-Experts model with 671B total parameters and 37B activated per token. Only the activated experts run on each forward pass, so inference cost tracks the 37B active count rather than the 671B total. Both checkpoints are MIT-licensed, which permits commercial use and self-hosting.

For a broader treatment of how context length affects cost and quality, see LLM context windows.

Benchmarks

Published scores place DeepSeek-V3.2-Exp in the top open tier for coding and reasoning. The table reports model-card figures. Treat cross-model comparisons cautiously, since harness and prompt differences move scores by several points.

The SWE-bench Verified jump from 49.2 (R1) to 67.8 (V3.2-Exp) is the headline for coding agents: a 18.6-point gain on real GitHub issue resolution between the January 2025 reasoning checkpoint and the November 2025 release. DeepSeek-R1 still posts strong pure-math numbers, with 97.3 on MATH-500 and 79.8 on AIME 2024.

For how these scores stack against other open models, see the best open-source coding model in 2026.

OpenAI-Compatible Calls

The DeepSeek API is a drop-in for the OpenAI SDK. Change base_url to https://api.deepseek.com, set your DeepSeek key, and pass deepseek-v4-flash or deepseek-v4-pro as the model. Streaming, function calling, and the message array work as they do against OpenAI.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_DEEPSEEK_API_KEY",
    base_url="https://api.deepseek.com",  # OpenAI-compatible endpoint
)

resp = client.chat.completions.create(
    model="deepseek-v4-flash",  # or "deepseek-v4-pro"
    messages=[
        {"role": "system", "content": "You are a senior Python engineer."},
        {"role": "user", "content": "Write a retry decorator with exponential backoff."},
    ],
)

print(resp.choices[0].message.content)

The same shape works from cURL. Set the Authorization header to your key and POST the OpenAI-style body to the chat completions path.

curl https://api.deepseek.com/chat/completions \
  -H "Authorization: Bearer YOUR_DEEPSEEK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {"role": "user", "content": "Explain MoE routing in two sentences."}
    ]
  }'

Because the contract is OpenAI-shaped, a single OpenAI-compatible client can target DeepSeek, OpenAI, or Morph's router by swapping the base URL. Morph's router at api.morphllm.com follows the same format, so the call below changes only the host and model and adds automatic difficulty-based routing.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_MORPH_API_KEY",
    base_url="https://api.morphllm.com/v1",  # OpenAI-compatible router
)

# The router classifies difficulty (~430ms) and routes easy calls
# to cheaper models, cutting cost 40-70% across the request mix.
resp = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Rename this variable for clarity."}],
)

How DeepSeek Pricing Compares

DeepSeek's value is the combination of a 1M-token context and per-1M prices well under most hosted frontier APIs. The table below sets deepseek-v4-flash and deepseek-v4-pro against two other open-weight APIs from the same facts pack to anchor the range. Closed-model comparisons are omitted where verified numbers were not available.

On output price, deepseek-v4-flash at $0.28/1M undercuts MiniMax-M2 ($1.20/1M) and GLM-4.6 ($2.20/1M) by 4x to 8x, while carrying a larger 1M-token context. deepseek-v4-pro at $0.87/1M output still sits below both. The cache-hit input rate ($0.0028/1M on flash) widens the gap further on repeated-prefix workloads.

The tradeoff is operational, not headline price. A single-provider API means a single point of rate limits, region, and capacity. Routing across providers through one endpoint hedges that. See LLM cost optimization for the full set of levers (caching, routing, batching, model tiering).

Getting an API Key

Create a DeepSeek platform account, open the API keys section, and generate a key. Set it as the api_key in the OpenAI SDK or as an Authorization Bearer header, then point base_url at https://api.deepseek.com. Billing is usage-based at the per-1M-token prices above, metered on input (split by cache state) and output.

If you want one key that reaches many models, Morph issues a single key at morphllm.com that works against the OpenAI-compatible router at api.morphllm.com, including the DeepSeek-class dsv4flash. That removes per-provider key management when you call more than one model family.

Frequently Asked Questions

How much does the DeepSeek API cost?

deepseek-v4-flash costs $0.14/1M input on a cache miss, $0.0028/1M on a cache hit, and $0.28/1M output. deepseek-v4-pro costs $0.435/1M input (cache miss), $0.003625/1M (cache hit), and $0.87/1M output. Both have a 1M-token context. Prices are from DeepSeek's pricing page as of June 2026.

What is the difference between deepseek-chat and deepseek-reasoner?

deepseek-chat is non-thinking (direct answer) mode and deepseek-reasoner is thinking (chain-of-thought) mode of deepseek-v4-flash. Both are legacy aliases that deprecate on 2026/07/24 at 15:59 UTC. After that you call deepseek-v4-flash and toggle thinking in the request.

What is the DeepSeek context window?

The deepseek-v4-flash and deepseek-v4-pro API models both expose a 1M-token context window. The open-weight checkpoints are smaller: DeepSeek-V3.2-Exp supports 163,840 tokens (160K) and DeepSeek-R1 supports 128K.

Is the DeepSeek API OpenAI-compatible?

Yes. Point the OpenAI SDK at base_url https://api.deepseek.com, set your DeepSeek key, and call client.chat.completions.create with model deepseek-v4-flash or deepseek-v4-pro. No code changes beyond the base URL and model name.

DeepSeek V4 vs R1: which should I use?

deepseek-v4-flash and deepseek-v4-pro are the current API models with a 1M context and split cache pricing. DeepSeek-R1 (671B total, 37B activated, 128K context, 49.2 on SWE-bench Verified) is the earlier January 2025 reasoning checkpoint. V4 supersedes R1 for new builds; R1 weights stay available under MIT for self-hosting.

How do I get a DeepSeek API key?

Create a DeepSeek platform account, generate a key in the API keys section, set it as the api_key in the OpenAI SDK, and point base_url at https://api.deepseek.com. Billing is usage-based per the prices above.

Can I reach a DeepSeek-class model through one shared endpoint?

Yes. Morph's OpenAI-compatible router at api.morphllm.com serves many models through one endpoint, including dsv4flash, a DeepSeek-class model on Morph's fleet. The router classifies difficulty in ~430ms and routes easy calls to cheaper models for 40-70% savings.

Related Resources

• DeepSeek V4: Specs, Context, and Benchmarks
• What Is an LLM Router? Automatic Model Routing
• LLM Cost Optimization
• LLM API Comparison
• LLM Context Windows Explained
• Best Open-Source Coding Model 2026
• Anthropic API Pricing
• LLM Cost Calculator